← All organizations

Generative AI models and inference optimization

Pruna AI

Pruna AI develops generative models and provides serverless inference for developers and creative teams. Its P-Image family supports image generation and editing, while P-Video supports video generation from text, images, and audio, with a draft mode for previewing concepts before full-quality output. Related tools handle virtual garment try-on, upscaling, and talking avatars. Alongside these hosted products, its open-source optimization framework lets developers compress and evaluate their own models; Pruna Pro adds advanced compression algorithms through the same interface.

The company’s four cofounders are CEO Rayan Nait Mazi, president and chief scientist Bertrand Charpentier, CTO John Rachwan, and chief strategy officer Stephan Günnemann. Its engineering approach brings pruning, quantization, distillation, and caching into a common framework, allowing developers to combine techniques rather than integrate each separately. Evaluation agents assess efficiency and output quality after compression. Pruna opened the framework’s core in 2025, extending its optimization tools to developers working within their existing machine learning workflows.

By March 2025, the company reported publishing more than 10,000 compressed AI models on Hugging Face. Pruna also raised a $6.5 million seed round in 2024 led by EQT Ventures, with participation from Daphni, Motier, and Kima, to expand its technical and commercial teams.

www.pruna.ai

1 talk

Newest first

1 speaker at AIE

Affiliations reflect their AIE appearances, not necessarily current employment.

Messages from the stage

Selecting models along a Pareto front

Charpentier advocates assessing specialized, compressed models by balancing task-specific quality against latency, cost, and efficiency. This makes model selection a question of application requirements and acceptable tradeoffs.

Affiliations reflect each recorded session, not necessarily current employment.

Company sources · checked 2026-08-28